Comparison of bias resulting from two methods of self-reporting height and weight: a validation study

OBJECTIVES: To contrast the validity of two modes of self-reported height and weight data. DESIGN: Subjects’ self-reported height and weight by mailed survey without expectation of subsequent measurement. Subjects were later offered a physical exam, where they self-reported their height and weight again, just prior to measurement. Regression equations to predict actual from self-reported body mass index (BMI) were fitted for both sets of self-reported values. Residual analyses assessed bias resulting from application of each regression equation to the alternative mode of self-report. Analyses were stratified by gender. SETTING: Upstate New York. PARTICIPANTS: Subjects (n = 260) with survey, pre-exam and measured BMI. MAIN OUTCOME MEASURES: Prevalence of obesity based on two modes of self-report and also measured values. Bias resulting from misapplication of correction equations. RESULTS: Accurate prediction of measured BMI was possible for both self-report modes for men (R (2 )= 0.89 survey, 0.85 pre-exam) and women (R (2 )= 0.92 survey, 0.97 pre-exam). Underreporting of BMI was greater for survey than pre-exam but only significantly so in women. Obesity prevalence was significantly underestimated by 10.9% (p < 0.001) and 14.9% (p < 0.001) for men and 5.4% (p = 0.007) and 11.2% (p < 0.001) for women, for pre-exam and survey, respectively. Residual analyses showed that significant bias results when a regression model derived from one mode of self-report is used to correct BMI values estimated from the alternative mode. CONCLUSIONS: Both modes significantly underestimated obesity prevalence. Underestimation of actual BMI is greater for survey than pre-exam self-report for both genders, indicating that equations adjusting for self-report bias must be matched to the self-report mode.